np.nan incorrectly casted into datetime on powerpc, leading to failing tests of pandas #8325
Description
originally detected while building fresh pandas for debian [1], and then verified with current master of numpy:
(sid_powerpc-dchroot)yoh@partch:~/numpy$ python -c 'import numpy as np; print(np.array([np.nan]).astype("M8[ns]")); print np.__version__'; git describe --tags
['2262-04-10T00:12:45.292707840']
1.13.0.dev0+a4dca24
pre-removal-numpybook-3581-ga4dca24expected result (e.g. on amd64 box):
$> python -c 'import numpy as np; print(np.array([np.nan]).astype("M8[ns]")); print np.__version__';
['NaT']
1.11.2[1] https://buildd.debian.org/status/fetch.php?pkg=pandas&arch=powerpc&ver=0.19.1-2&stamp=1480352140
edit 1: issue is not new (debian jessie/stable):
(jessie_powerpc-dchroot)yoh@partch:~$ python -c 'import numpy as np; print(np.array([np.nan]).astype("M8[ns]")); print np.__version__'
['2262-04-10T00:12:45.292707840+0000']
1.8.2